AMD - Piledriver chitchat

[watercooled]

I guess it shows the importance of some sort of minimum measurements; the subjective analysis confirms it here, but otherwise it's IMO important to have a way of filtering out rare low spikes e.g. like those 'time spent beyond ...ms' charts, provided they're done properly of course.

Just looking at average measurements, you can see why people would pick something like a Pentium for certain games, but then end up with a rubbish experience when it stutters unbearably when a lot is happening.

[watercooled]

Tom's have done some direct power measurements of the 295X2 with their new kit: http://www.tomshardware.com/reviews/...e,3799-16.html

[watercooled]

There may currently be BIOS issues with some of those new Gigabyte Bay Trail motherboards, if anyone was thinking of buying one, especially for *nix: https://forum.pfsense.org/index.php?...0511#msg410511

[CAT-THE-FIFTH]

It seems Intel has lost over $900 million this quarter with the part of the company which handles Atom.

[CAT-THE-FIFTH]

AMD presentation on die stacking:

http://www.microarch.org/micro46/files/keynote1.pdf

[Goobley]

AMD presentation on die stacking:

http://www.microarch.org/micro46/files/keynote1.pdf

Wow... as someone who spent a few weeks attempting (unsuccessfully) to etch single MOSFETs onto silicon wafer I am always amazed at what the industry is going to mass produce. Whilst I haven't looked through the document in full detail yet my initial impression is that this could only be a good thing for reducing productions costs as single dies get smaller and then have to be joined - hopefully higher yield as well as rendering new chips feasible? Interesting stuff, I'll read it more fully tomorrow when I get chance.

[CAT-THE-FIFTH]

Overclocking the AM1 CPUs:

http://www.phoronix.com/scan.php?pag...clocking&num=1
http://forums.anandtech.com/showthread.php?t=2378163

It seems adjustments are limited and it means you cannot get a massive overclock.

Edit!!

It seems in the 2nd review,there is quite a decent overclock with the Athlon 5150.

[Noxvayl]

AMD presentation on die stacking:

http://www.microarch.org/micro46/files/keynote1.pdf

That looks very interesting, I look forward to buying everything on one chip for a media PC. I hope that instead of an SSD they either have a second layer of CPUs and keep a CPU only design without a graphics card for the high end desktop parts, would be great to have 16 cores for an enthusiast CPU.

[scaryjim]

It seems in the 2nd review,there is quite a decent overclock with the Athlon 5150.

I suspect it's down to ratio dividers in the BIOS - I remember with the original Llano APUs bclock overclocking would have stable ranges where the mobo manufacturers put in dividers to knock the peripheral clocks back down to ~ 100MHz. Based on the stable clocks in that thread (125MHz and 133MHz) it looks like ASUS might have done the same thing (in which case there might also be one around 150MHz, I guess...)

[DanceswithUnix]

I presume these parts will use the resonant mesh technology, which might give odd effects like that.

[Bonebreaker777]

It seems in the 2nd review,there is quite a decent overclock with the Athlon 5150.

I would like see the max.performance squeezed out at stock frequency plus memory OC.

Does these boards for AM1 have facilities for memory OC (like Z series Intel chipsets only) or the OC happens because they increase the bclock?

In the review they said the heatsink uses a different mounting that the sockets in the previous generations.
Does this mean that cheap second hand coolers designed for chips with a lot more TPD are not a viable option-even going as far as removing the FAN and running in passive mode?

ECC support haven't been mentioned so far and AMD chips usually support it, but it was down to the board manufacturer to enable support.It is the same with AM1 as well?

[watercooled]

The AM1 cooler mounting system uses two pushpins, but IIRC there are boards available that use the 115x mounting system instead: http://www.techpowerup.com/199223/ec...therboard.html

Although as noted in that link, the contact height may differ, so not all coolers are guaranteed to work.

[DanceswithUnix]

Memory overclock on AM1 is mentioned by AMD so seems quite common.

[Bonebreaker777]

The AM1 cooler mounting system uses two pushpins, but IIRC there are boards available that use the 115x mounting system instead: http://www.techpowerup.com/199223/ec...therboard.html

Although as noted in that link, the contact height may differ, so not all coolers are guaranteed to work.

Ahh.How's the noise produced by the stock coolers-even the temperatures long term without a FAN in an average or good ventilated case?

Memory overclock on AM1 is mentioned by AMD so seems quite common.

Least something.Wish memory overclock for me too.

Is undervoltage,while maintaining stock speed,a possibility-as for a machine intended to be on 24/7 while reducing consumption (undervoltage CPU,undervoltage and underclock the IGP,etc)???

[watercooled]

Well underclocking/volting doesn't make much difference at idle for modern CPUs as power gating is used to completely switch off inactive parts. It can have an effect on load power consumption, but the extent you'll be able to do it depends on the individual chip, like overclocking. And it's just as important to test stability thoroughly.

[Bonebreaker777]

I can see a very effective Home Server with undervoltaged CPU (but still stable) and an additional SATA card.
How's the ECC support?